Networks cover a wide range of connectivity. There are Wide Area Networks (WANs), such as the Internet, Local Area Networks (LANs), such as within a small office, and micro-networks (μNetworks), such as on an integrated circuit (IC) chip.
The term μNetwork is used to describe on-chip networks that are used to connect intellectual property (IP) cores such as processors, digital signal processors (DSPs), memory systems, and peripherals into a functioning computer system. It is beneficial for such a system to have well-defined error handling properties. In particular, in some systems it may be required that parts of the system can continue unaffected by errors in other parts of the system, and that such errors may be corrected and the system brought back to full functionality without requiring a full system shut-down or reset. The present invention deals with error handling in networks and μNetworks. The present invention discloses partitioning the task between hardware and software, and what specific hardware mechanisms may be needed in a μNetwork in order to allow error isolation and recovery.